accessibility
Pass
Audited by Gen Agent Trust Hub on Sep 4, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill is primarily educational, offering patterns and checklists for web accessibility. It includes an auditing agent that uses read-only tools to scan for accessibility violations. No evidence of malicious intent, obfuscation, or credential harvesting was found.
- [INDIRECT_PROMPT_INJECTION]: The a11y-auditor agent (in agents/a11y-auditor.md) has a potential surface for indirect prompt injection as it processes project files. (1) Ingestion points: The agent reads HTML, JSX, CSS, and JS files from the project directory using the Read and Grep tools. (2) Boundary markers: No specific delimiters or warnings to ignore instructions within the ingested code are present. (3) Capability inventory: The agent is restricted to Read, Glob, and Grep tools; it has no network or write access. (4) Sanitization: No sanitization of the input file content is performed. Assessment: This is a low-risk surface inherent to its primary purpose as an auditing tool and does not escalate the overall verdict.
Audit Metadata